Denitrifying Bioreactorsâ€”An Approach for Reducing Nitrate Loads

G Model ECOENG-1647; No. of Pages 12 ARTICLE IN PRESS

Ecological Engineering xxx (2010) xxx–xxx

Contents lists available at ScienceDirect

Ecological Engineering

journal homepage: www.elsevier.com/locate/ecoleng

Review Denitrifying bioreactors—An approach for reducing nitrate loads to receiving waters

Louis A. Schipper a,∗, Will D. Robertson b, Arthur J. Gold c, Dan B. Jaynes d, Stewart C. Cameron e a Department of Earth and Ocean Sciences, University of Waikato, Private Bag 3105, Hamilton, New Zealand b Department of Earth and Environmental Sciences, University of Waterloo, Waterloo, ON N2L 3G1, Canada c Department of Natural Resources Science, University of Rhode Island, Kingston, RI 02881, USA d National Laboratory for Agriculture and the Environment, USDA-ARS, 2110 University Blvd, Ames, IA 50011-3120, USA e GNS Science, Private Bag 2000, Taupo, New Zealand article info abstract

Article history: Low-cost and simple technologies are needed to reduce watershed export of excess nitrogen to sensitive Received 12 November 2009 aquatic ecosystems. Denitrifying bioreactors are an approach where solid carbon substrates are added Received in revised form 9 March 2010 into the flow path of contaminated water. These carbon (C) substrates (often fragmented wood-products) Accepted 3 April 2010 act as a C and energy source to support denitrification; the conversion of nitrate (NO −) to nitrogen gases. Available online xxx 3 Here, we summarize the different designs of denitrifying bioreactors that use a solid C substrate, their hydrological connections, effectiveness, and factors that limit their performance. The main denitrifying Keywords: bioreactors are: denitrification walls (intercepting shallow groundwater), denitrifying beds (intercepting Denitrification concentrated discharges) and denitrifying layers (intercepting soil leachate). Both denitrifcation walls and Bioreactor − Nitrate beds have proven successful in appropriate field settings with NO3 removal rates generally ranging from −3 −1 −3 −1 Effluent 0.01 to 3.6gNm day for walls and 2–22gNm day for beds, with the lower rates often associated with nitrate-limitations. Nitrate removal is also limited by the rate of C supply from degrading substrate − and removal is operationally zero-order with respect to NO3 concentration primarily because the inputs − − of NO3 into studied bioreactors have been generally high. In bioreactors where NO3 is not fully depleted, removal rates generally increase with increasing temperature. Nitrate removal has been supported for up to 15 years without further maintenance or C supplementation because wood chips degrade sufficiently slowly under anoxic conditions. There have been few field-based comparisons of alternative C − substrates to increase NO3 removal rates but laboratory trials suggest that some alternatives could sup- − port greater rates of NO3 removal (e.g., corn cobs and wheat straw). Denitrifying bioreactors may have a number of adverse effects, such as production of nitrous oxide and leaching of dissolved organic matter (usually only for the first few months after construction and start-up). The relatively small amount of field data suggests that these problems can be adequately managed or minimized. An initial cost/benefit analysis demonstrates that denitrifying bioreactors are cost effective and complementary to other agricultural management practices aimed at decreasing nitrogen loads to surface waters. We conclude with recommendations for further research to enhance performance of denitrifying bioreactors. © 2010 Elsevier B.V. All rights reserved.

1. Introduction trated discharges in agricultural systems occur through under-ﬁeld tile drainage and ditches (Duda and Johnson, 1985; Dinnes et al., The nitrogen (N) cascade is an increasingly important global 2002) while diffuse sources are typically through the discharge of issue as N ﬂowing through ecosystems has multiple impacts on shallow groundwater to surface waters. Nitrogen that is captured in terrestrial, aquatic and atmospheric environments (Galloway et al., biomass passes through the food chain and ends up in wastewater 2003, 2008). In agricultural systems, N in excess of plant and ani- streams, which are ultimately discharged to surface waters. mal needs can leach to shallow groundwater and ultimately enter A number of strategies are being implemented to reduce the N surface waters through concentrated or diffuse discharges. Concen- load to aquatic ecosystems. Sophisticated technologies have been developed to remove N from wastewater and are employed in municipal treatment plants and septic tank systems (Oakley et ∗ Corresponding author. Tel.: +64 7 858 3700; fax: +64 7 858 4964. al., this issue). In agricultural ecosystems, many land manage- E-mail address: [email protected] (L.A. Schipper). ment approaches have been proposed to reduce N losses, such as

Please cite this article in press as: Schipper, L.A., et al., Denitrifying bioreactors—An approach for reducing nitrate loads to receiving waters. Ecol. Eng. (2010), doi:10.1016/j.ecoleng.2010.04.008 G Model ECOENG-1647; No. of Pages 12 ARTICLE IN PRESS

2 L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx improved N use efficiency in crops, managing N inputs through bioreactors (Figs. 1 and 2; Tables 1 and 2). Designs differ primarily − soil tests, land management plans and controlled drainage (Sims in the hydrologic connection between water containing NO3 and et al., 1995; Drury et al., 1996; Dinnes et al., 2002; Jaynes et C source and the ratio of source area:treatment area. Broadly, we al., 2004). Integrating wetland and riparian buffers into the agri- use terminology first used by Robertson and Cherry (1995). cultural landscape have also been demonstrated as a potentially “Denitrification walls” are where solid C material is incorpo- useful way to reduce N losses to surface waters (Hill, 1996; Kadlec, rated vertically into shallow groundwater perpendicular to the 2005). flow. Darcian flow, saturated hydraulic conductivity, hydraulic gra- The most widely understood process of permanent N removal dient and the flow paths intercepted by the walls controls the flux − from terrestrial and aquatic ecosystems is heterotrophic denitri- of NO3 into the walls. Walls may intercept natural groundwater − fication, which converts nitrate (NO3 ) to N gases using a carbon flow paths, or groundwater flow paths that have been altered by (C) source as the electron donor and for growth (Seitzinger et al., subsurface tile drainage systems or by the morphometry and rela- 2006; Coyne, 2008; Rivett et al., 2008). In this review, we focus on tively higher saturated hydraulic conductivity of C additions within heterotrophic denitrification but we recognise that other micro- the wall. The source area is roughly limited to the boundaries of the bial processes such as anaerobic ammonium oxidation (Anammox) wall orthogonal to the flow direction. While the term “wall” sug- and chemo-autotrophic denitrification can also produce N gases gests a barrier to flow, these walls are designed to sustain elevated (Burgin and Hamilton, 2007). The rates and controlling factors of hydraulic conductivities (i.e. >10 m day−1) conducive to substan- these microbial processes deserve further attention, but are not tial rates of shallow groundwater flow. Denitrification walls can covered in this review. be 100% wood chips (Farhner, 2002; Jaynes et al., 2008) or saw- At the microbial scale, the rate of heterotrophic denitrifica- dust mixed with soil (Robertson and Cherry, 1995; Schipper and − tion is controlled by concentrations of oxygen (O2), NO3 and C Vojvodic-Vukovic, 1998). (Seitzinger et al., 2006). The availability of a degradable C source “Denitrification beds” are containers (sometimes lined) that − − to drive denitrification becomes critical where NO3 is present are filled with wood chips and receive NO3 in concentrated dis- in excess, such as in many wastewater plants and agricultural charges either from a range of wastewaters (Robertson et al., settings. Aerobic microorganisms obtain energy through the oxi- 2005a; Schipper et al., this issue) or tile/drain discharge (Blowes dation of organic compounds, using O2 as the electron acceptor, et al., 1994; Robertson et al., 2009; Robertson and Merkley, until the environment becomes energetically favourable for the 2009). Denitrification beds have also been installed into existing − use of NO3 as an electron acceptor. As such, organic C plays two stream beds or drainage ditches and are specifically referred to key roles in promoting denitrification; firstly, to provide an anoxic as “stream bed bioreactors” (Robertson and Merkley, 2009). The environment and, secondly, to act as an electron donor for deni- source area:treatment area ratio for beds is usually much greater trification. In wastewater treatment systems, this C is part of the than in wall designs, due to either natural or artificial drainage waste stream, but can also be supplemented with liquid C sources networks that intercept and funnel groundwater inputs to the such as methanol (Oakley et al., this issue; Henze et al., 2008). In bioreactor. Beds, referred to as “upflow bioreactors” have also been agricultural soils, denitrification can be limited by sufficient labile adapted for use along streambanks (van Driel et al., 2006a,b). These C to create an anoxic environment and provide energy for denitri- upflow bioreactors do not receive input from a specific point source, fication. but have design features and placement requirements that induce Here, we review passive technologies that have been recently focused flowpaths that resemble the high flux conditions associ- developed to overcome the C limitation of denitrification for ated with point sources. Upflow bed reactors rely on wood chips − enhanced NO3 removal. A variety of carbonaceous solids and with high saturated hydraulic conductivity and create a favourable immiscible liquids have been successfully tested in denitrifying hydraulic gradient by lowering the water table within the bed bioreactors (Hunter, 2005; Gibert et al., 2008), although many through the placement of a drainage pipe near the top of the wall of C sources have only been tested in laboratory trials. To date, that discharges directly to the adjacent stream. wood-particle media in particular, has been the most widely used Finally, “denitrification layers” are horizontal layers of solid C material in field trials and has shown an ability to deliver consistent material that have been installed under weeping tiles from sep- − longer term (5–15 years) NO3 removal, while requiring mini- tic tank drainage fields (Robertson and Cherry, 1995) or under mum maintenance (Robertson et al., 2000, 2008, 2009; Schipper effluent-irrigated topsoils (Schipper and McGill, 2008). and Vojvodic-Vukovic, 2001; Fahrner, 2002; Schipper et al., 2005; The key to selecting an appropriate bioreactors design depends Jaynes et al., 2008). Consequently, this review has a strong focus on the hydrologic conditions and site constraints of the system of on the use of wood-particle media, although other carbonaceous interest (Table 2, and see Section 4). solids, including some with higher reaction rates, could also play an important role as additional experience is gained. Our focus is on field trials of solid C substrates that have been used in agri- 3. Removal rates and controlling factors cultural and rural settings to enhance denitrification and decrease − discharges of NO3 from either diffuse discharges (e.g., shal- 3.1. Removal rates low groundwater) or concentrated discharges (e.g., effluents, tile drainage, small streams and ditches). Because these approaches Nitrate removal rates have been reported for a wide range of are intended to be applied throughout the landscape at the scale of denitrifying bioreactors (Table 3) using different units. For consis- − individual fields, septic systems and tributary streams it is impor- tency, in this synthesis paper, NO3 removal rates are expressed −3 tant that the technologies are simple, passive and that maintenance asgNO3-N removed per m reactor volume per day (g NO3- requirements are minimal. Nm−3 day−1). Where conversion of rate units was required for study comparisons (Table 3), an effective porosity value of 0.7 (van Driel et al., 2006a) was used in most cases. It is also possible to − 2. Terminology express NO3 removal rates based on surface area of the bioreactor; this can be useful when wanting to compare the performance − There are multiple designs that use solid C sources for enhanc- of bioreactors to NO3 removal in wetlands or other terrestrial ing denitrification and are collectively referred to as denitrifying ecosystems (for example, where effluent is applied to land).

L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx 3

− Fig. 1. Schematic of different designs of denitrification walls: (A) plan view of a wall intercepting NO3 plume, (B) side view of wall installed to impervious layer forcing shallow groundwater through wall, (C) wall not installed to full depth to impervious layer allowing flow paths under, (D) high permeability layer installed in upper part of − aquifer causing upwelling of groundwater containing NO3 , and (E) denitrification walls installed on either side of a subsurface tile drain.

− Sustained NO3 removal rates for denitrification beds incorpo- because walls can have wood mixed with inert material such as soil rating wood, range from about 2 to 22gNm−3 day−1. Variation in or sand. The highest reported removal rate (15gNm−3 day−1) was rate is predominantly attributed to bed operating temperatures for a 100% sawdust wall constructed in Western Australia (Fahrner, ◦ − − (typically from 2 to 20 C) and/or influent NO3 concentrations 2002), which received very high NO3 concentrations and had rela- − (see fuller discussion below). The highest sustained NO3 removal tively high soil temperatures (Table 3); all of which would promote rates were measured by Blowes et al. (1994) and Robertson et relatively high denitrification rates. al. (2000) in a denitrification bed (North Campus, Canada) using We have not addressed the importance of the dominant denitri- woodchips. Removal rates for this trial varied between 4 and fying organisms in the denitrifying bioreactors but these organisms 22gNm−3 day−1 depending on temperature (2–20 ◦C). Other den- are very wide-spread (Coyne, 2008), and bioreactors studied to date itrification bed studies (van Driel et al., 2006a,b; Schipper et al., have not required inoculation as the denitrifiers respond rapidly this issue) utilizing a combination of sawdust and woodchip media to environmental drivers. However, microbial population diversity − report a slightly lower rate of NO3 removal, varying between 2 and and dynamics deserve further attention as appropriate molecular 20gNm−3 day−1. Robertson et al. (2005a) measured lower removal tools are developed (Wallenstein et al., 2006). rates (2–5gNm−3 day−1), during evaluation of the commercially available NitrexTM filter which contains a mixture of sawdust and 3.2. Nitrate concentration woodchips; however, removal rates were nitrate-limited at these sites. While denitrification is likely to obey Michelis–Menton kinetics − Nitrate removal rates for denitrification walls (Robertson with regard to NO3 concentration, most denitrifying bioreactors − et al., 2000; Schipper et al., 2005; Jaynes et al., 2008) con- receive NO3 concentrations higher than Km of denitrifying bac- taining wood media are generally an order of magnitude teria (see Barton et al., 1999). Consequently, when considering −3 −1 − lower (0.014–3.6gNm day ) than denitrification beds. Nitro- whether NO3 removal follows either zero-order and first-order gen removal rates within walls may be limited by low rates of kinetics, the situation may be viewed as functionally similar to − − NO3 loading, as most walls removed virtually all the NO3 but also zero-order kinetics in many cases (Robertson et al., 2000; Schipper

4 L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx

Fig. 2. Schematic of different designs of denitrification beds: (A) side view of bed treating concentrated discharges of effluent or drainage water, (B) plan view of bed installed to intercept drainage water from agricultural land, (C) plan view of bed installed to intercept drainage water leaching from septic tank drainage fields, and (D) bed installed into base of stream. et al., 2005), although first order approaches have also been used generally because wood is commonly available at low cost, sup- in some studies (Chun et al., 2009; Leverenz et al., this issue). ports high permeability, has a high C:N ratio (ranging from 30:1 − There is experimental evidence to support the notion that NO3 to 300:1 depending on source and type of wood material; Gibert removal in denitrifying bioreactors is operationally zero-order. In et al., 2008; Vogan, 1993) and long durability (Robertson et al., a series of column tests using aged woodchip media (fresh to 7 2009). There have been a number of smaller-scale laboratory exper- − years old), successive runs at increasing influent NO3-N concen- iments, which have examined the NO3 removal rates in a range −1 − trations (3.1–49 mg L ) did not result in increased NO3 removal of other C substrates (Table 4 and see also Gibert et al., 2008). − −3 (Robertson, this issue). This indicated that the reaction was con- Some of the highest reported NO3 removal rates (19–105gNm trolled by an independent parameter (presumably the release rate d−1) are from the column studies of Vogan (1993) and Shao et al. of degradable C from the carbonaceous media) and thus zero- (2009) using cellulose, alfalfa, wheat straw, and rice husk as the C order reaction kinetics would apply over this concentration range. substrate. − A transition to first-order kinetics may occur at lower NO3 con- While more labile C sources (e.g., cracked corn, corn stalks, centrations (such as might be found in stormwater), but in most straw, etc.) may support higher removal rates than wood media, − settings, zero-order kinetics likely represents much of the NO3 these may require more frequent replenishment because of rapid transformation, and could be used for most design purposes. C depletion. For example, Stewart et al. (1979) found that humus − rich soil was ineffective for long-term NO3 removal from sep- 3.3. Alternative C sources tic tank effluent due to relatively rapid depletion of available C. Decreases in the saturated hydraulic conductivityof the more labile Field-scale denitrification bed and wall studies have mainly C sources may also occur as the C structure breaks down. Cameron used wood products (sawdust and wood chips) as a C source, and Schipper (this issue) found that while maize cob media sup-

L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx 5

Table 1 Physical description and potential settings for different denitrifying bioreactor designs.

Nomenclature Physical description Objective Potential settings Field examples

− Denitrification wall (i) C substrate placed into the Removal of NO3 from Down gradient of localized Canada (Robertson et al., upper 1–2 m of shallow ground water prior to sources of nitrate-enriched 2000), New Zealand groundwater in a trench surface water recharge groundwater; at focal (Schipper et al., 2005), perpendicular to groundwater points of groundwater Australia (Fahrner, 2002) flow path towards surface flow. water. − (ii) C substrate placed into the Removal of NO3 from In subsurface drained Jaynes et al. (2008) upper 1–2 m of shallow groundwater before agricultural land groundwater in a trench on entering drainage either side of a tile drain network − Denitrification bed Container (varied length and Removal of NO3 from Concentrated discharges van Driel et al. (2006a) and − breadth dimensions but wastewaters or tile that have high NO3 Schipper et al. (this issue) typically 1–2 m deep) filled drainage from concentrations such as with solid C substrate. Effluent agricultural fields. from tiles or treated or drainage water enters and wastewater exits in pipes. Beds may be lined. − Upflow bioreactors (subset of C substrate in container with Removal of NO3 from Adjacent to surface water van Driel et al. (2006b) denitrification beds) lined sides, open to ground water prior to where groundwater is groundwater flow at bottom. surface water recharge shallow and aquifers have Groundwater flows towards C lower conductivities than substrate with elevated added C substrate saturated hydraulic conductivity and is discharged to adjacent stream via pipe. − Stream-bed bioreactor Container (varied length and Reducing NO3 Streams and drainage Robertson and Merkley (subset of denitrification breadth dimensions but concentrations in ditches (2009) beds) typically 1–2 m deep) filled streams with solid C substrate installed in the base of a stream − Denitrification layer A horizontal layer of Reduce NO3 leaching Below septic wastewater Canada (Robertson et al., woodchips that receives vertically to drainage field that passed 2000) New Zealand nitrified effluent from above. groundwater through a sand/gravel filter (Schipper and McGill, or other land-based 2008) effluent treatment system

− ported a 6.5-fold higher NO3 removal rate than wood media over available in some areas and may be more expensive. Both hard- a 24-month period, the decline in saturated hydraulic conductivity wood and softwood species have been used successfully in field was generally greater in the maize cob media. trials (Robertson et al., 2000; Schipper and Vojvodic-Vukovic, 1998; The effect of media particle size on reaction rates has also van Driel et al., 2006a,b; Jaynes et al., 2008) with little advantage been considered. Several studies have found no significant dif- indicated of one over the other. − ference in the NO3 removal rates in wood-particle media of While laboratory-scale experiments provide opportunity for − different particle sizes (Carmichael, 1994; van Driel et al., 2006a,b; relative comparison of the NO3 removal potential and hydraulic Robertson et al., 2000; Cameron and Schipper, this issue), although performance of different C substrates, these studies may not be − Greenan et al. (2006) measured initially greater NO3 removal a reliable indication of removal rates or hydraulic performance for both wood chips and cardboard when ground to <2 mm. In achievable in larger-scale field installations. This is due in part to comparing two denitrification beds, one constructed with predom- the effect of dissolved O2 (DO) content of the influent water on inantly coarse-grained wood particles (woodchips, 1–50 mm) and removal rates in small scale trials. Also laboratory trials tend to be − the other constructed with fine grained wood particles (sawdust, of short duration, typically less than 6 months and NO3 removal − 1–5 mm), van Driel et al. (2006a) measured NO3 removal rates rates tend to decline with time as labile C is reduced (Schipper of 5.9gNm−3 day−1 for coarse grained media and a similar rate and Vojvodic-Vukovic, 1998). The results of short-term experi- of5.5gNm−3 day−1 for fine grained media. Robertson et al. (2000) ments may not be reliable for assessing longer term sustainablity speculated that denitrification is associated with reaction rims that of removal rates (Cameron and Schipper, this issue). Multiyear penetrate, by diffusion, into the carbonaceous solids, rather than field testing of wood media in bioreactors has shown that rates being restricted to the grain surfaces. This was supported by exam- measured after about 1 year of operation are generally representa- ination of a denitrification bed after 4 years of operation, which tive on long-term operation (Robertson et al., 2000; Schipper and revealed that the larger centimeter-sized wood particles exhibited Vojvodic-Vukovic, 2001; Jaynes et al., 2008). Recommended best dark coloured rims that extended several millimetres into the par- practice would be field testing of favourable substrates identified ticles, while the centre of the particles remained light coloured and from laboratory studies for a minimum of a year. Ultimately the appeared unaltered from their original condition. They interpreted selection of an appropriate C media is a balance between availabil- this to indicate that the dark rims represented the denitrification ity, cost and reaction rate. zone. The smaller woodchips were dark coloured throughout. − There is no evidence of a difference in the NO3 removal rate 3.4. Temperature achieved by hardwood in comparison to softer, faster growing tree species (Gibert et al., 2008; Cameron and Schipper, this issue). It In general, biological reactions rates increase with increasing is possible that hardwood may maintain physical properties for temperature. Examination of reaction rates from a variety of stud- longer due to greater density. However, hardwood is not as readily − ies where NO3 was non-limiting (Table 3) support a general

6 L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx

Table 2 Hydrological connections, limitation and potential approaches to overcoming limitations in denitrifying bioreactors.

Nomenclature Hydrological connections Limitations Potential for overcoming limitations

− Denitrification wall (i) Down-gradient of NO3 source. Flow Requires site-specific analyses of In aquifers>2mdeep, upwelling of − into wall is governed by Darcian principles hydraulic gradient, the depth and NO3 laden groundwater into the − of groundwater flow. Pore water velocities extent of NO3 plumes. Removal of interceptor wall may be enhanced by −1 − are likely to be low (0.05–0.5 m day ) and NO3 limited to up-gradient source increasing the width (parallel to flow retention times within the wall are areas and within the upper 1–2 m of path) of the interceptor wall. expected to be 3–10 days. Retention time groundwater. If wall has lower within the wall is governed by incoming saturated hydraulic conductivity − flow rate (passive). Hydrologic properties than surrounding aquifer, NO3 are likely to display high spatial (i.e., K) and plumes may flow under or around temporal (i.e., hydraulic gradient) the wall. variability. (ii) For walls adjacent to tile drains, For walls adjacent to tile drains, high − groundwater flow moves through flows with high NO3 concentrations bioreactors following drainage flow lines. lead to short retention times − Retention times are limited when water resulting in large losses of NO3 . tables are highest and flow is greatest. Denitrification bed For tile drains: Flow rate is governed by Seasonality of flow rates can create High flow bypass can be incorporated properties of subsurface drainage network high flow situations with limited into design to minimize flooding and including: area of drained land, pattern retention rates within the bed, overflow within the ditch. Cellular − and extent of excess rainfall or irrigation, limiting NO3 removal. designs can be coupled with flow designed depth of water table decrease diverters to optimize retention times due to drainage system, and intensity of at different flow regimes. drainage network. Retention time within the bed is governed by incoming flow rate. For wastewater: flows controlled by upstream waste generation and wastewater plant management. Upflow bioreactor Groundwater upwells into open bottom of Very site specific applications. ‘Rip-rap’ can be used to control reactor: (i) C substrate has higher Stream side location may be prone to erosion conductivity than the aquifer. (ii) Drainage erosion; flow rate is effected by pipes installed at upper portion of reactor stream stage discharges to adjacent stream, lowers water table within reactor and enhances hydraulic gradient to upflow reactor. Stream-bed bioreactor A bioreactor installed in bottom of stream Seasonal flows over the top of the Rip rap cover and channel narrowing with a down-gradient riffle creates a riffle resulting in partial treatment. can minimize siltation pressure gradient and stream water flows Potential for siltation of surface inlet down through lower conductivity C requiring cleaning substrate and through an exit pipe back into the lower reach of the stream Denitrification layer Loading is determined by the design of the Extent of nitrification in preceding Ensure appropriate pre-treatment, septic system. sand/gravel filter can limit N e.g., sand filter removal. Difficult to replace C substrate because underneath discharge.

Table 3 − − Rates of NO3 removal for a range denitrifying bioreactors in the ﬁeld. In general average rates of NO3 removal are presented. Many of the systems recorded here had − −3 −1 −3 complete NO3 removal which would limit the rate of denitriﬁcation and consequently are likely underestimates of potential removal rate. Units are g N m d where m refers to volume of bioreactor.

− System design Study Size of bioreactor Typical NO3 inputs Temperature annual Average rate of N removal (m3) (gNm−3) average (◦C) (gNm−3 d−1)

Walls Robertson et al. (2000) 1 50 14 1.7 Schipper et al. (2005) 78 5–15 12 1.4a Jaynes et al. (2008) 79 87 10 0.62 Fahrner (2002) 160 >60 19 12.7

Beds Robertson et al. (2000) 2 5 10 10 Robertson et al. (2005a,b) 9 17 15 1.8a 108 38 15 2.4a 120 35 15 2.5a 360 14 15 5.1a van Driel et al. (2006a) upﬂow reactors 0.7 9 9 2.1 0.2 13 13 3.7 Robertson and Merkley (2009) 40 5 8 3.2 Robertson et al., 2009 17 10 7.7 3.4 Schipper et al. (this issue) 83 53 15–25 1.4a 294 5.5 0–11a 1320 250 20 9.7

Layer Robertson et al. (2000) 1 57 10 1.8

a − Nitrate removal rate limited by NO3 concentrations.

L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx 7

Table 4 Table 5 Laboratory-scale experiments (column and batch) that have tested a range of C Theoretical percent of dissolved organic C (DOC) that is utilized for denitrification − sources for NO3 removal. when dissolved O2 (DO) reduction occurs first. Calculation uses the stoichiome- try of Robertson and Merkley (2009) for the denitrification and aerobic respiration Carbon media Reference reactions, and assumes initial DO of 8 mg L−1 (saturation value at 25 ◦C). Wood Vogan (1993), Carmichael (1994), Healy et NO -N removed (g m−3) DOC utilized for denitrification (%) al. (2006), Greenan et al. (2006), Gibert et 3 al. (2008) and Cameron and Schipper (this 122 issue) 346 Cardboard Greenan et al. (2006) 10 74 Newspaper Volokita et al. (1996a,b) 30 90 Wheat straw Vogan (1993), Soares and Abeliovich (1998) and Cameron and Schipper (this issue) 3gNm−3, either because of short retention times (less than sev- Alfalfa Vogan (1993) eral hours) or because of low NO − concentrations, consumption Corn Fay (1982) 3 Maize cobs Cameron and Schipper (this issue) of available C by aerobes can exceed that by denitrifying bacteria. − Soil Stewart et al. (1979) and Gibert et al. Initially poor NO3 removal in the denitrification bed field trial (2008) of Healy et al. (2006) was attributed to high DO concentration Soil and sawdust Healy et al. (2006) (3.7–7.3 mg L−1) and the relatively short retention time. Soil and jute pellets Wakatsuki et al. (1993) Sulfate can also be present in wastewaters and groundwaters Cotton Volokita et al. (1996b), Della Rocca et al. (2005, 2006), Su and Puls (2007) and act as an alternative electron acceptor when more reduc- Compost, mulch or greenwaste Gibert et al. (2008) ing conditions develop. However, denitrifying organisms generally Cameron and Schipper (this issue) out-compete sulfate reducers for available C (Appelo and Postma, Seaweed Ovez et al. (2006) 1994), consequently sulfate reduction normally only occurs when − NO3 concentrations have been substantially depleted (Vogan,

− 1993; Robertson and Merkley, 2009; Robertson et al., 2009; positive relationship between NO3 removal and average annual Robertson, this issue; Woli et al., this issue; Elgood et al., this issue). temperature (Robertson et al., 2008, 2009; Cameron and Schipper, Sulfate-reducing conditions are often also accompanied by increas- this issue). Field testing of near-surface reactors in Canada has − ing DOC concentrations (Vogan, 1993; Robertson, this issue). Thus, shown that these continue to remove NO3 at a modest rate reactors that have excessively long retention times, beyond that (∼2gNm−3 day−1) even at effluent temperatures as low as 1–5 ◦C − required to fully deplete NO3 , risk generating high levels of DOC (Robertson and Merkley, 2009; Robertson et al., 2009; Elgood et al., and undesirable reaction by-products such as hydrogen sulfide. If this issue). In the latter study, the stream-bed bioreactor continued − sulfate reduction is significant there is the possibility of production to operate and provided NO3 removal throughout the winter sea- of toxic methyl mercury (Woli et al., this issue) by sulfate reducing son even though the stream surface was periodically frozen. These bacteria; however, this production has yet to be measured. studies were conducted using a range of wood-particle materials and it is likely that other factors also influenced the temperature 4. Hydrology response such as the degradability of different C substrates. For example, in a 23-month barrel study, Cameron and Schipper (this ◦ The application of denitrifying bioreactors requires an under- issue) found that Q10 (the increase in rate for a 10 C increase − standing of the specific hydrologic settings and the spatial and in temperature) for NO3 removal differed between C substrates − ◦ temporal patterns of NO3 flux at the site (Table 2). Consequently, ranging from less than 0.8 to 2.3, between 14 and 24 C. In the case site investigations differ between walls, beds and layers and are of maize cobs, Cameron and Schipper (this issue) found that in the − site-specific. longer term NO3 removal was less at the higher temperature, pre- − For denitrification walls, the NO3 flux relies on both Darcian sumably because labile C had been depleted more rapidly at higher − − principles and the extent of groundwater NO3 contamination. If a temperature. While NO3 removal rate generally increases with − denitrification wall is located in an area with either low NO3 con- increasing temperature, further information is needed for different centrations or low groundwater flow rates, the removal rates will substrates for design purposes. − be quite low due to NO3 limitations. Hence, a number of aquifer characteristics must be investigated to determine site suitability 3.5. Processes competing for available C and design parameters, including:

− • A key determinant for NO3 removal is the availability of C to Depth to the water table. Low cost wall construction usually pre- denitrifiers and any microbial process that out-competes denitri- cludes placing C material deeper than 4–5 m depth, therefore the − fiers for this C will reduce NO3 removal in denitrifying bioreactors. water table needs to be within 2–3 m of the ground surface in Dissolved O2 in either groundwater or in point source discharges most cases. may allow aerobes to out-compete denitrifiers for available C • Pattern and values of saturated hydraulic conductivity. The wall (Rivett et al., 2008). This is most likely an issue when retention should be placed in media that is conducive to groundwater flow times are short but less likely of concern in large bioreactors with (K >1mday−1). long retention. Laboratory column tests have shown that the time • Depth to a restrictive layer with low saturated hydraulic conduc- required to deplete the dissolved O2 in DO saturated water is tivity. approximately 1 h in aged, 2-year-old woodchip media (Robertson, • Direction and gradient of groundwater flow. In many areas, this issue) and field trials have indicated similar rates of DO removal hydraulic gradient undergoes marked seasonal changes, thus site in wood particle reactors (Down, 2001; Robertson et al., 2009). investigations should target both high and low water table con- Data in Table 5 shows the percent of dissolved organic C (DOC) ditions. • − that would theoretically be consumed by denitrification and aero- Spatial pattern of shallow groundwater NO3 concentrations. − − bic respiration in DO saturated water, for a range of NO3 removal Where walls are to be placed near localized sources of NO3 amounts. In bioreactors where NO3-N removal is less than about input, such as septic systems or animal waste disposal sites, the

8 L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx

− potential for narrow plumes warrants field investigations with denitrification is the dominant mechanism of NO3 removal in den- 15 − multiple monitoring wells to optimize site location. itrifying bioreactors. Greenan et al. (2006) added N-labelled NO3 into wood chip columns and laboratory incubations and found that immobilisation and DNRA accounted for less than 4% of total NO − Bioreactor designs that treat concentrated flows (e.g., tile drains 3 removed. Similarly, Gibert et al. (2008) concluded that less than or wastewater) or surface flows, avoid the necessity for site specific − studies of shallow groundwater flow conditions, and therefore can 10% of NO3 removed was attributable to DNRA (and generally have a considerable cost advantage in many cases. this was less than 5%). Further evidence against the occurrence + Differences between the saturated hydraulic conductivity of of DNRA is the lack of significant ammonium (NH4 ) produc- the denitrifying bioreactor and the surrounding aquifer can either tion observed in most denitrification walls and beds (Schipper − and Vojvodic-Vukovic, 1998; Robertson et al., 2000, 2005a, 2007; diminish or enhance NO3 flux into the wall. Schipper et al. (2004) Schipper et al., this issue). Although, there is little evidence for found that in situ mixing of sawdust into saturated sands sub- DNRA occurring in wood particle barriers, it could be more impor- stantially lowered the saturated hydraulic conductivity of the wall tant when considering other more labile carbonaceous substrates. compared to the coarse sand media that composed the shallow Vogan (1993) observed an NH + −1 aquifer. Consequently, some shallow groundwater bypassed below 4 increase of 5 mg N L in a lab- the wall, substantially decreasing overall NO − removal. Barkle oratory column test utilizing alfalfa straw as a C source. Because 3 + − the NH4 increase occurred in the same column zone where NO3 et al. (2008) demonstrated that mixing of sands that were satu- − rated while constructing a denitrification wall caused a decline in was being depleted and not further along the column where NO3 saturated hydraulic conductivity, presumably because of particle was fully depleted and DOC was still increasing, it was concluded + resorting. Furthermore, they showed that the addition of larger par- that the NH4 increase was probably the result of DNRA, rather ticles of organic material (garden mulch) did not improve saturated than from mineralization of organic N from the media. Long- hydraulic conductivity when mixed with sand. However, walls term accumulation of N into organic matter is difficult to assess composed entirely of wood chips can have very high hydraulic because even large amounts of N accumulation in added C source conductivities in excess of 100 m day−1 and can induce ground- would only result in small (and likely undetectable) declines in C:N water flow convergence and upwelling particularly as the width ratio. of the wall in the direction of flow is increased (Robertson et al., In support of active denitrification in bioreactors, elevated lev- 2005b). Several studies have used model simulations to provide els of denitrifying enzyme activity (DEA) have commonly been insight into the nature of groundwater flow in and around reactive measured in denitrification walls and layers (Schipper et al., 2004, barriers considering a variety of geometries and permeability char- 2005; Schipper and McGill, 2008; Barkle et al., 2008; Moorman acteristics (Benner et al., 2001; Robertson et al., 2005b, 2007). These et al., this issue). However, elevated DEA does not always mean that significant denitrification is occurring, Schipper and McGill studies were also accompanied by field tests with detailed moni- − toring in order to validate model simulations. Nitrate flux into walls (2008) did not measure much NO3 removal in a denitrification can also be enhanced by their placement at locations where ambi- layer despite increased DEA. In this case, leaching rates were very high and residence times short (<1 day). Further evidence to sup- ent hydraulic gradients have been enhanced. For example Jaynes − et al. (2008), placed denitrification walls parallel to tile drains in port denitrification as a major mechanism for NO3 removal comes from a field study conducted by Schipper and Vojvodic-Vukovic artificially drained fields. Current studies have been mostly small − (2000) who calculated NO3 removal in a denitrification wall using scale and more work is required to better understand the nature − of groundwater flow in and around reactive barriers particularly as hydraulic gradients, saturated conductivity and NO3 concentra- they are scaled up in size and a wider range of configurations and tions and compared these calculations to laboratory measures of media types are tested. denitrification rate. Schipper and Vojvodic-Vukovic (2000) found that denitrification rates (0.6–18.1 ng N cm−3 h−1) were sufficiently Design of denitrification beds requires information on the sea- − sonal variability in flow rate and NO − flux. Bypass systems that high to account for NO3 removal from shallow groundwater 3 −3 −1 divert flows around a denitrification bed can be installed where (0.8–12.8 ng N cm h ); however, error bars for both measure- ment approaches were large. A subsequent study at the same site, flow rates vary greatly or are flashy with rainfall. − where NO3 was injected into the denitrification wall, measured In many cases it may be useful to adopt a mass removal − approach when considering bioreactor use. If the selected biore- greater rates of NO3 removal than could be accounted for by lab- − oratory measurements of denitrification (Schipper et al., 2005). actor design can provide adequate flows and NO3 levels to avoid − Moorman et al. (this issue) measured denitrifcation potential in NO3 -limiting conditions, then the amount of nitrate removal can be easily estimated simply by considering the volume of media uti- bioreactors to be comparable to activity in the surface 15 cm of lized and the published reaction rates (Section 3.1). This approach an organic rich field soil. In contrast to the soil, enrichment with could become very attractive if (when) programs of nutrient trad- glucose did not increase denitrification potential in the bioreactor ing are adopted. With trading incentives, bioreactors would be indicating that the woodchips provided sufficient C substrate for preferentially deployed at the most efficient, lowest cost loca- the denitrifiers. tions, for example in shallow sand and gravel aquifers with high Another line of evidence supporting denitrification activity is 15 − groundwater fluxes in intensively cultivated landscapes with high enrichment of N in the residual NO3 , because denitrifying bacteria preferentially consume the lighter isotope (14N). In contrast, groundwater NO3 concentrations. − immobilisation of NO3 into organic pools does not discriminate between isotopes (Mariotti et al., 1982). In laboratory studies using 5. Mechanism for nitrate removal woodchips of varying age (fresh to 7-year-old), Robertson (this − 15 issue) observed progressive enrichment of NO3 - N as depletion In most studies of denitrifying bioreactors, it is assumed that proceeded. Enrichment was similar in all four woodchip types and conventional heterotrophic denitrification is the dominant mech- was consistent with a Rayleigh-type distillation process with iso- − − anism of NO3 removal. Other possible fates for NO3 include N topic fractionation factor of −13 per mil. Similar fractionation has immobilisation into organic matter, dissimilatory nitrate reduc- been reported for denitrification in groundwater (Robertson, this − 15 tion to ammonium (DNRA), and Anammox (Burgin and Hamilton, issue). Enrichment of NO3 - N has also been observed in field 2007). There is evidence in both laboratory and field studies that bioreactors (Down, 2001; Robertson et al., 2000) In each of these

L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx 9

+ −1 −1 −1 studies NH4 concentrations remained low (<1 mg N L ) making 33 mg L declined to <10 mg L after 10 days of column oper- −1 it unlikely that DNRA or Anammox caused the enrichments. ation. However, substantial NO3-N concentrations (>10 mg L ) remained in the column effluent which may have assisted with DOC consumption. Likewise, Carmichael (1994) observed DOC deple- 6. Potential adverse effects of using denitrifying tion to <15 mg L−1 after 25 pore volumes in a laboratory column bioreactors utilizing 100% woodchips with a 2-day retention time. Interest- ing, DOC declined from 43 mg L−1 at pore volume 22 down to A general concern associated with enhancing denitrification 11 mg L−1 at pore volume 26, coincident with the breakthrough − is that nitrous oxide (N2O), a greenhouse gas, can also be pro- of NO3 in the column effluent. Thus, at sites where high initial duced as a by-product, but so far there are only a few studies DOC concentrations may be unacceptable, control measures might of N2O fluxes from denitrifying bioreactors. In column studies, include maintaining high flow rates during start-up, installing post- Greenan et al. (2009) measured N2O emissions that were between bioreactor treatment (e.g., sandfilter) or collection of the initial − 0.003 and 0.028% of the total NO3 removed. This was much effluent for disposal elsewhere. In agricultural terrain, a practi- less than IPCC default value for production of N2O assumed to cal solution may be to use the initial effluent water for irrigating − ultimately arise from NO3 leaching from soil (0.75%; Mosier et adjacent fields. Another option is to pre-leach the media prior to al., 1998). N2O fluxes were also measured from a denitrifica- use, but this is likely to be logistically difficult adding to costs and tion wall and adjacent pasture using closed chamber techniques has not, as yet, been attempted. Robertson and Merkley (2009) for 2-year period and fluxes were significantly greater (P < 0.05) observed that slightly elevated concentrations of total phenolic − − from the wall (average 0.31gNha 1 h 1) than the adjacent pas- compounds (4 ␮gL−1) persisted in the Avon streambed reactor into − − ture (average 0.05gNha 1 h 1) (Schipper and Vojvodic-Vukovic, its second year of operation. Similarly, water passing through a unpublished data). It is likely that N2O emissions will be low- denitrifying bioreactor will have DO removed by microbes and this est from denitrifying bioreactors with complete or near-complete deoxygenated water could have adverse effects on biota of receiv- − NO3 removal. This was the case for N2O flux from a stream-bed ing waters. Currently, there is little information on this potential reactor (Avon site, Canada, Elgood et al., this issue) where dis- impact, but presumably this would depend on the relative flow − solved N2O was found to be much lower when NO3 removal rates of the reactor and the receiving waters and the nature of the −1 − was complete (0–5 ␮gNL ) compared to when NO3 removal receiving waters. −1 was incomplete (10–35 ␮gNL ). Overall dissolved N2O produc- − tion amounted to 0.5% of NO3 removed over a 1-year period. Similarly, Moorman et al. (this issue) estimated N2O production 7. Longevity of nitrate removal and maintenance of − at 0.62% of NO3 removed in denitrification walls on either side of saturated hydraulic conductivity a tile drain. − Other gases will also be released from the denitrifying bioreac- The longevity of NO3 removal in denitrifying bioreactors is tors including CO2 and potentially CH4 both derived from decaying not fully known because currently there appear to be no examples organic matter. The emission of CO2 does not result in a net increase of reactors that have failed due to C depletion (Robertson et al., in CO2 emissions as the C substrate used in the bioreactor would 2008). Two factors will affect longevity—the continued supply of C have decayed if used for other purposes. Elgood et al. (this issue) to denitrifiers and the maintenance of adequate saturated hydraulic collected gas bubbles erupting from the surface of the Avon stream- conductivity. Decomposition of solid C sources (e.g., woodchips, bed reactor, and found these to contain substantial amounts of CH4 sawdust) are greatly slowed when water saturated conditions are (25–45%). Methane was also detected during early operation of the maintained and, in most cases, only slow rates of C decomposition − bioreactors described by Jaynes et al. (2008), but disappeared after are needed to support NO3 removal because there is generally a − a few months presumably as highly labile C in the wood chips was large amount of C relative to NO3 inputs. consumed. Theoretically, the fluxes of CH4 from bioreactors should Currently, the most long-lived bioreactor is the denitrification − be low when NO3 concentrations remain sufficiently high to sup- wall of Robertson and Cherry (1995), constructed in 1992 to treat press methanogens; however, this concept requires validation. a septic system plume. In a recent re-examination of this wall During start-up of denitrifying bioreactors there is also the in 2007 (year 15, Robertson et al., 2008), core samples of the − potential for the release of soluble C compounds (measured as reactive media provided a NO3 removal rate in laboratory tests, biochemical oxygen demand, BOD) associated with fresh wood of ∼4gNm−3 day−1, which was only about 50% lower than the material or other solid C sources. Release of BOD can reduce DO rate measured in year 1 (7gNm−3 day−1). Schipper et al. (2005) − in receiving waters and adversely affect biota. Fresh wood con- reported ongoing NO3 removal in a denitrification wall over 7 tains 1–2 wt% soluble organic constituents such as tannic acids years of operation, and found that denitrification continued to be (Vogan, 1993), which can rapidly leach from wood-particle media nitrate-limited, rather than C-limited, throughout this period. In during start-up. Initial reactor effluent is normally dark coloured the denitrification walls constructed by Jaynes et al. (2008), more −1 − and can have DOC concentrations of hundreds of mg L . This is than 60% of the NO3 was removed during the first 2 years of oper- similar to the leachate that occurs at sawmills and in log storage ation while removal was slightly more than 50% on average in the yards where control measures focus on high DOC, trace metals and following 6 years (Moorman et al., this issue). Moorman et al. (this phenolic compound levels (Taylor et al., 1996). In wood particle issue) measured an exponential decrease in the C content over the reactors, the duration and magnitude of soluble DOC leaching dur- first 8 years for wood chips located at the saturated/unsaturated ing start-up is dependant upon the reactor retention time. Several interface of the bioreactor. After 8 years, only about 25% of the C field studies have observed dissipation of the initial DOC spike over content of these woodchips remained with an estimated half life of the first 3–6 months of reactor operation (Robertson and Cherry, 4.6 years. However, for woodchips that were deeper in the bioreac- 1995; Robertson et al., 2005b; Leverenz et al., this issue; Schipper et tor and below the water table for a greater fraction of the year, more al., this issue). In other reactor studies with shorter retention times, than 80% of the C still remained (estimated half life of 36.6 years), − more rapid dissipation of the initial DOC spike has been observed. which accounted for the continued NO3 removal efficacy of the In column tests with sand and sawdust media and one day reten- bioreactor. In a comparative study of woodchip media of varying tion time, Vogan (1993) observed that the initial effluent DOC of age, Robertson (this issue) measured reaction rates in core samples

10 L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx

Table 6 9. Treatment of other contaminants Cost/benefit analysis of N removal of a denitrifying bioreactor (Robertson et al., 2009) installed for treatment of tile drainage in comparison to other agricultural approaches for reducing N. The majority of data on the performance of denitrifying biore- − actors has logically focused on NO3 removal but there are −1 Practice US$ (kg-N ) Source other redox sensitive contaminants in wastewaters and shallow Bioreactor 2.39–15.17 This paper groundwater that might be treated using modified bioreac- Soil testing and side dressing N 1.15 Saleh et al. (2007) tors. These include pathogens and pharmaceutical compounds in fertilizer wastewaters, pesticides in agricultural drainage and industrial con- Drainage water management 2.71 Jaynes and Thorp (2008) Wetlands 3.26 Hyberg (2007) taminants such as perchlorate which is associated with explosives Fall cover crops 11.06 Saleh et al. (2007) and rocket fuel manufacturing. Monitoring of four sawdust beds treating septic tank effluent in Ontario (Robertson et al., 2005a),

− showed that over several years of operation, Escherichia coli lev- from two reactors that were 2 and 7 years old, and found that NO3 els generally remained below detection in the reactor effluents removal rates remained within 50–75% of rates measured in fresh (<10 cfu 100 mL−1). However, several breakouts did occur, up to woodchips. ∼1000 cfu 100 mL−1, particularly at the site with the highest load- With such slow degradation rates, woodchip bioreactors have ing rate. Robertson et al. (2007) reported complete attenuation demonstrated an ability to remain highly permeable over a number of trace levels of perchlorate (ClO4) occurring in agriculturally of years with no deterioration in saturated hydraulic conductiv- impacted groundwater in Ontario, during migration through a ∼ −1 ity ( 100 m day ) evident (Robertson et al., 2009). However, the woodchip layer. In a subsequent experiment, Robertson et al. bioreactor in the latter study was a surface installation that was (2009) immersed a highway safety flare containing ClO4 in the inlet subject to minimal overburden pressures. Subsurface bioreactors pipe of a woodchip reactor treating agricultural drainage and the that experience greater overburden pressures could potentially be −1 elevated influent ClO4 concentration (up to 33 ␮gL ) was entirely prone to greater deterioration in saturated hydraulic conductivity attenuated in the reactor. It was noted that ClO4 attenuation did with time. However, little long-term saturated hydraulic conduc- − not commence until NO3 was first depleted, demonstrating that tivity data is currently available for subsurface bioreactors. − NO3 inhibits the degradation of ClO4. − + − − Of the N forms (NO3 ,NH4 , organic N), only NO3 and NO2 8. Cost–benefit analysis seem to be removed with little, if any, removal of organic N or ammonium (Robertson et al., 2005a; Schipper et al., this issue). − While it is clear that denitrifying bioreactors can remove NO3 This reinforces the need for a nitrification step of wastewater prior from concentrated or diffuse discharges, there are other technolo- to being loaded into a denitrifying bioreactor. gies/approaches that are also effective. To be applied in the “real Phosphate is often at low concentrations in groundwater but can world”, denitrifying bioreactors need to be cost effective when be significant in effluents. To date, there has been little evidence − compared to these other approaches for managing NO3 , e.g., that wood-based bioreactors remove phosphate from effluents installing wetland and riparian buffers, or more intensive manage- (Robertson et al., 2005a; Jaynes et al., 2008; Schipper et al., this ment of N application. issue). However, the addition of other amendments to bioreactors To estimate the cost effectiveness of bioreactors, the bed reactor (e.g., iron slag), has resulted in considerable phosphate removal described in Robertson et al. (2009) will be used as an example. from treated septic tank effluent (Baker et al., 1998; Robertson, This bioreactor was 13 m long × 1.2 m wide × 1.1 m deep for a total 2000) and streams (McDowell et al., 2008). 3 −1 volume of 17.2 m . It removed an average of 11.3 kg N year from Emerging contaminants such as pharmaceutical compounds an agricultural field drain. Assuming a conservative 20-year life also have the potential to be attenuated in denitrifying bioreactors expectancy for this bioreactor, a total of 226 kg N will be removed. but monitoring data is, as yet, lacking. In the central US, woodchips can be purchased for US$ 26.50 m−3. Hauling to a site within 35 km would cost an additional US$ 65. Rental of a backhoe for installation would range from US$ 500 to 10. Conclusions and future work 1000 and incidental expenses would add another US$ 50. Assuming 4% annual interest for the “time value of money” for the installation To date, field studies have demonstrated that denitrifying biore- − costs gives a total cost of installation of US$ 3249 or a cost per actors are capable of substantial NO3 removal in a number of − −1 removal of NO3 of US$ 15.17 kg-N . watershed settings. Major advantages of denitrifying bioreactors However, many bioreactors might be installed on farms where are their simplicity with low maintenance requirements and the farmers have access to both a backhoe and wood from wind breaks ability to tailor designs to fit hydrological site criteria. It is not clear − or other local sources, greatly reducing the cost of installation. In how long these systems will continue to remove NO3 because addition, the bioreactor described in Robertson et al. (2009) only no studied systems has yet been observed to fail—consequently − removed NO3 for about 70% of the year when the field tile was we can only conclude that denitrifying bioreactors could last for a − draining. Connecting a bioreactor to a NO3 source that flowed year minimum of 15 years (Robertson et al., 2008). Moorman et al. (this − round would increase its efficiency. Thus, for a farmer installing issue) suggest that sustainability of wood chips to support NO3 − this bioreactor onto a year-round source the cost of NO3 removal removal was dependent on the time that wood chips remained −1 would be reduced to US$ 2.39 kg-N . This cost of removal range water saturated (and presumably anoxic). Half lives varied between (US$ 2.39–15.17) compares favourably with estimates of other 4.6 and 36.6 years for woodchip either periodically saturated or − NO3 removal technologies as shown in Table 6. Cost efficiencies permanently saturated, respectively. − − for bioreactors of other designs would of course vary, but biore- Where NO3 loading is high, NO3 removal is dependent on − − actors can be cost efficient alternatives for removing NO3 when temperature and availability of C. The high loads of NO3 enter- − compared against other commonly promoted approaches for man- ing denitrifying bioreactors mean that often NO3 removal rates + aging N. In many cases, denitrifying bioreactors are complementary are zero-order. Organic N and NH4 are not removed in denitri- with these other practices and do not preclude the use of multiple fying bioreactors and a nitrification step may be required before mitigation approaches (Woli et al., this issue). wastewater enters the bioreactor. Development of simple nitrifi-

L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx 11 cation pre-treatment systems coupled with bioreactors should be Cameron, S.C., Schipper, L.A., this issue. Nitrate removal and hydraulic perfor- encouraged such as those described by Oakley et al. (this issue). mance of carbon substrates for potential use in denitrification beds. Ecol. Eng., doi:10.1016/j.ecoleng.2010.03.010. There is some laboratory and field evidence that denitrifica- Carmichael, P.A., 1994. Using Wood Chips as a Source of Organic Carbon in Denitri- − tion is the main mechanism for NO3 removal but further work fication: A Column Experiment and Field study Implementing the Funnel and on the microbial ecology of these systems is needed and may lead gate Design. M.Sc. Thesis. Dept Earth Sci., University of Waterloo, Waterloo, ON, Canada. to approaches for increasing nitrate removals rates and longevity. Chun, J.A., Cooke, R.A., Eheart, J.W., Kang, M.S., 2009. Estimation of flow and transport Field-scale research is needed to determine the effectiveness, costs, parameters for woodchip-based bioreactors: 1. Laboratory-scale bioreactor. − and factors controlling the rate of NO3 removal and denitrification Biosyst. Eng. 104, 384–395. in different bioreactors, particularly the suitability of alternative Coyne, M.S., 2008. Biological denitrification. In: Schepers, J.S., Raun, W. (Eds.), Nitrogen in Agricultural Systems. ASA-CSSSA-SSSA Agronomy Monograph 49, C sources. Management approaches still need to be developed Madison, WI, pp. 197–249. for decreasing unwanted side effects, such as the production of Della Rocca, C., Belgiorno, V., Meric, S., 2005. Cotton-supported heterotrophic den- N O and initial leaching of dissolved organic matter. Design cri- itrification of nitrate-rich drinking water with sand filtration post treatment. 2 Water SA 31, 1022–1028. teria and demonstration sites are warranted to test alternative Della Rocca, C., Belgiorno, V., Meriç , S., 2006. Heterotrophic/autotrophic denitrifi- designs that merge bioreactors with constructed wetlands to pro- cation (HAD) of drinking water: prospective use for permeable reactive barrier. vide co-benefits of biodiversity and aesthetics (Leverenz et al., Desalination 210, 194–204. Dinnes, D.L., Karlen, D.L., Jaynes, D.B., Kaspar, T.C., Hatfield, J.L., Colvin, T.S., Cam- this issue). Integration of bioreactors with other approaches for bardella, C.A., 2002. Nitrogen management strategies to reduce nitrate leaching − reducing NO3 loads to surface water, such as riparian zones or in tile-drained Midwestern soils. Agron. J. 94, 153–171. controlled drainage, would also be beneficial. Down, J.E., 2001. Geochemical Reaction Distribution Within a NitexTM filter Treating Farm Field Drainage water. B.Sc. Thesis. Dept Earth Sci., University of Waterloo, While there are unanswered questions about performance of Waterloo, ON, Canada. denitrifying bioreactors, there is sufficient information available to Drury, C.F., Tan, C.S., Gayner, J.D., Olaya, T.O., Welacky, T.W., 1996. Influence of con- − trolled drainage-subsurface irrigation on surface and tile drainage nitrate loss. utilize bioreactors for reducing NO3 fluxes in a variety of settings. J. Environ. Qual. 25, 317–324. Design manuals should be developed that address site evalua- Duda, A.M., Johnson, R.J., 1985. Cost-effective targeting of agricultural nonpoint- tion, provide detailed construction approaches that integrate with source pollution controls. J. Soil Water Conserv. 40, 108–114. local hydrology while meeting policy directives and performance Elgood, Z., Robertson, W.D., Schiff, S.L., this issue. Greenhouse gas pro- goals of different countries and regions. These design manuals duction in a stream bed bioreactor for nitrate removal. J. Ecol. Eng., doi:10.1016/j.ecoleng.2010.03.011. might be developed for engineers but, where appropriate, also Fahrner, S., 2002. Groundwater Nitrate Removal using a Bioremediation Trench. for farmers and farm advisors. Finally, of particular importance Honours Thesis. University of Western Australia, Perth. is determining linkages between hydrological flow paths, reten- Fay, L.D., 1982. Simulation of Biological Denitrification in Columns Representing − Recirculating Sand Filters. BSc. Thesis. Univ of Michigan. tion time in the bioreactors and NO3 removal efficiency, thus Galloway, J.N., Aber, J.D., Erisman, J.W., Seitzinger, S.P., Howarth, R.W., Cowing, E.B., we recommend interdisciplinary research combining the skills of Cosby, B.J., 2003. The nitrogen cascade. Bioscience 53, 341–356. hydrologists, hydrogeologists, engineers, biogeochemists, and land Galloway, J.N., Townsend, A.R., Erisman, J.W., Bekunda, M., Cai, Z.C., Freney, J.R., Martinelli, L.A., Seitzinger, S.P., Sutton, M.A., 2008. Transformation of the nitro- managers. gen cycle: recent trends, questions, and potential solutions. Science 320, 889– 892. Gibert, O., Pomierny, S., Rowe, I., Kalin, R.M., 2008. Selection of organic substrates Acknowledgments as potential reactive materials for use in a denitrification permeable reactive barrier (PRB). Bioresour. Technol. 99, 7587–7596. This paper is a product of a workshop on “Denitrification in Man- Greenan, C.M., Moorman, T.B., Kaspar, T.C., Parkin, T.B., Jaynes, D.B., 2006. Comparing carbon substrates for denitrification of subsurface drainage water. J. Environ. aged Ecosystems” held May 12–14, 2009, at the University of Rhode Qual. 35, 824–829. island Bay Campus, Narragansett, RI, with support from the Deni- Greenan, C.M., Moorman, T.B., Kaspar, T.C., Parkin, T.B., Jaynes, D.B., 2009. Denitri- trification Research Coordination Network of the National Science fication in wood chip bioreactors at different water flows. J. Environ. Qual. 38, Foundation, award DEB0443439 and the USDA CSREES Northeast 1664–1671. Healy, M.G., Rodgers, M., Mulqueen, J., 2006. Denitrification of a nitrate-rich syn- States and Caribbean Islands Regional Water Project award 2008- thetic wastewater using various wood-based media materials. J. Environ. Sci. 51130-19504. Eric Davidson is thanked for initiating the workshop. Health Part A 41, 779–788. This work was partially supported by USDA NRCS-RI agreement Henze, M., van Loosdrecht, M.C.M., Ekama, G.A., 2008. Biological Wastewater Treatment: Principles, Modeling, and Design. Intl Water Assn Publishing, number 69-1535-06-005, Landcare Research contract C09X0705 528 pp. University of Waikato, GNS Science. We thank Bob Kadlec, Ken Hill, A.R., 1996. Nitrate removal in stream riparian zones. J. Environ. Qual. 25, Foreman, and Kelly Addy for early discussions on the focus of this 743–755. Hunter, W.J., 2005. Injection of innocuous oils to create reactive barriers for biore- review. mediation: laboratory studies. J. Contam. Hydrol. 80, 31–48. Hyberg, S., 2007. Economics of CREP/CRP Treatment Wetlands for the tile Drained Cropland in the Corn Belt. Available at: http://www.fsa.usda.gov/ References Internet/FSA File/hyberg iowa wetlands.pdf. Jaynes, D.B., Thorp, K., 2008. Potential water quality impact of drainage water man- Appelo, C.A.J., Postma, D., 1994. Geochemistry, Groundwater and Pollution. A.A. agement in the Midwest cornbelt. In: ASABE Annual Meeting, Providence, RI. Balkema, Rotterdam, The Netherlands. Paper No. 084566, Jun 29–July 2. Baker, M.J., Blowes, D.W., Ptacek, C.J., 1998. Laboratory development of perme- Jaynes, D.B., Dinnes, D.L., Meck, D.W., Karlen, D.L., Camberella, C.A., Colvin, T.S., 2004. able reactive mixtures for the removal of phosphorus from onsite wastewater Using the late Spring nitrate test to reduce nitrate loss within a watershed. disposal systems. Environ. Sci. Technol. 32, 2308–2316. Agron. J. 33, 669–677. Barkle, G.F., Schipper, L.A., Burgess, C.P., Milne, J., Painter, B.D., 2008. Insitu mixing of Jaynes, D.B., Kaspar, T.C., Moorman, T.B., Parkin, T.B., 2008. In situ bioreactors and organic matter decreases conductivity of denitrification walls in sand aquifers. deep drain-pipe installation to reduce nitrate losses in artificially drained fields. Ground Water Monit. Remediat. 28, 57–64. J. Environ. Qual. 37, 429–436. Barton, L., McLay, C.D.A., Schipper, L.A., Smith, C.T., 1999. Annual denitrification rates Kadlec, R.H., 2005. Nitrogen farming for pollution control. J. Environ. Sci. Health Part in agricultural and forested soils: a review. Aust. J. Soil Res. 37, 1073–1093. A: Environ. Sci. Eng. 40, 1307–1330. Benner, S.G., Blowes, D.W., Molson, J.W.H., 2001. Modeling preferential flow in Leverenz, H.L., Haunschild, K., Hopes, Tchobanoglous, G., Darby J. this reactive barriers; implications for performance and design. Ground Water 39, issue. Anoxic treatment wetlands for denitrification. Ecol Eng., 371–379. doi:10.1016/j.ecoleng.2010.03.014. Blowes, D.W., Robertson, W.D., Ptacek, C.J., Merkley, C., 1994. Removal of agricultural Mariotti, A., Mariotti, F., Champigny, M.L., Amarger, N., 1982. Nitrogen isotope frac- − nitrate from tile-drainage effluent water using in-line bioreactors. J. Contamin. tionation associated with nitrate reductase activity and uptake of NO3 by Pearl Hydrol. 15, 207–221. Millet. Plant Physiol. 69, 880–884. Burgin, A.J., Hamilton, S.K., 2007. Have we overemphasized the role of denitrifica- McDowell, R.W., Sharpley, A.N., Bourke, W., 2008. Treatment of drainage water with tion in aquatic ecosystems? A review of nitrate removal pathways. Front. Ecol. industrial by-products to prevent phosphorus loss from tile-drained land. J. Environ. 5, 89–96. Environ. Qual. 37, 1575–1582.

12 L.A. Schipper et al. / Ecological Engineering xxx (2010) xxx–xxx

Moorman, T.B., Parkin, T.B., Kaspar, T.C., Jaynes, D.B., this issue. Denitrification activ- Schipper, L.A., Vojvodic-Vukovic, M., 2001. Five years of nitrate removal, denitri- ity, wood loss, and N2O emissions over nine years from a wood chip bioreactor. fication and carbon dynamics in a denitrification wall. Water Res. 35, 3473– Ecol. Eng., doi:10.1016/j.ecoleng.2010.03.012. 3477. Mosier, A., Kroeze, C., Nevison, C., Oenema, O., Seitzinger, S., Van Cleemput, O., 1998. Schipper, L.A., Barkle, G.F., Vojvodic-Vukovic, M., Hadfield, J.C., Burgess, C.P., 2004. Closing the global N2O budget: nitrous oxide emissions through the agricul- Hydraulic constraints on the performance of a groundwater denitrification tural nitrogen cycle—ECD/IPCC/IEA phase ii development of IPCC guidelines for wall for nitrate removal from shallow groundwater. J. Contam. Hydrol. 69, national greenhouse gas inventory methodology. Nutr. Cycl. Agroecosyst. 52, 263–279. 225–248. Schipper, L.A., Barkle, G.F., Vojvodic-Vukovic, M., 2005. Maximum rates of nitrate Oakley, S., Gold, A.J., Oczkowski, A.J., this issue. Nitrogen control through decentral- removal in a denitrification wall. J. Environ. Qual. 34, 1270–1276. ized wastewater treatment: process performance and alternative management Schipper, L.A., Cameron, S., Warneke, S., this issue. Nitrate removal from strategies. Ecol. Eng., doi:10.1016/j.ecoleng.2010.04.030. three different effluents using large-scale denitrification. Ecol. Eng., Ovez, B., Mergaert, J., Saglam, M., 2006. Biological denitrification in drinking water doi:10.1016/j.ecoleng.2010.02.007. treatment using the seaweed Gracilaria verrucosa as carbon source and biofilm Seitzinger, S., Harrison, J.A., Bohlke, J.K., Bouwman, A.F., Lowrance, R., Peterson, B., carrier. Water Environ. Res. 78, 430–434. Tobias, C., Van Drecht, G., 2006. Denitrification across landscapes and water- Rivett, M.O., Buss, S.R., Morgan, P.P., Smith, J.W.N., Bemment, C.D., 2008. Nitrate scapes: a synthesis. Ecol. Appl., 2064–2090. attenuation in groundwater: a review of biogeochemical controlling processes. Shao, L., Xu, Z.X., Jin, W., Yin, H.L., 2009. Rice husk as carbon source and biofilm Water Res. 42, 4215–4232. carrier for water denitrification. Polish J. Environ. Stud. 18, 693–699. Robertson, W.D., 2000. Treatment of wastewater phosphate by reductive dissolution Sims, J.T., Vasilas, B.L., Gartley, K.L., Milliken, B., Green, V., 1995. Evaluation of soil of iron. J. Environ. Qual. 29, 1678–1685. and plant nitrogen tests for maize on manured soils of the Atlantic Coastal Plain. Robertson, W.D., this issue. Rates of nitrate removal in woodchip media of varying Agron. J. 87, 213–222. age. Ecol Eng., doi:10.1016/j.ecoleng.2010.01.008. Soares, M.I.M., Abeliovich, A., 1998. Wheat straw as a substrate for water denitrifi- Robertson, W.D., Cherry, J.A., 1995. In situ denitrification of septic-system nitrate cation. Water Res. 32, 3790–3794. using reactive porous media barriers: field trials. Ground Water 33, 99–111. Stewart, L.W., Carlile, B.C., Cassel, D.K., 1979. An evaluation of alternative simulated Robertson, W.D., Merkley, L.C., 2009. In-stream bioreactor for agricultural nitrate treatments of septic tank effluent. J. Environ. Qual. 8, 397–403. treatment. J. Environ. Qual. 38, 230–237. Su, C., Puls, R.W., 2007. Removal of added nitrate in the single, binary, and ternary Robertson, W.D., Blowes, D.W., Ptacek, C.J., Cherry, J.A., 2000. Long-term of perfor- systems of cotton burr compost, zerovalent iron, and sediment: implications mance of in situ reactive barriers for nitrate remediation. Ground Water 38, for groundwater nitrate remediation using permeable reactive barriers. Chemo- 689–695. sphere 67, 1653–1662. Robertson, W.D., Ford, G.I., Lombardo, P.S., 2005a. Wood-based filter for nitrate Taylor, B.R., Goudey, J.S., Carmichael, N.B., 1996. Toxicity of aspen wood leachate to removal in septic systems. Trans. ASAE 48, 121–128. aquatic life: laboratory studies. Environ. Toxicol. Chem. 15, 150–159. Robertson, W.D., Yeung, N., van Driel, P.W., Lombardo, P.S., 2005b. High- van Driel, P.W., Robertson, W.D., Merkley, L.C., 2006a. Denitrification of agricultural permeability layers for remediation of ground water; go wide, not deep. Ground drainage using wood-based reactors. Trans. ASAE 48, 121–128. Water 43, 574–581. van Driel, P.W., Robertson, W.D., Merkley, L.C., 2006b. Upflow reactors for riparian Robertson, W.D., Ptacek, C.J., Brown, S.J., 2007. Aquifer nitrate and perchlorate reme- zone denitrification. J. Environ. Qual. 35, 412–420. diation using a wood particle layer. Ground Water Monit. Remediat. 27, 85–95. Vogan, J.L., 1993. The use of Emplaced Denitrifying Layers to Promote Nitrate Robertson, W.D., Vogan, J.L., Lombardo, P.S., 2008. Nitrate removal rates in a 15-year Removal from Septic Effluent. M.Sc. Thesis. Dept Earth Sci., University of Water- old permeable reactive barrier treating septic system nitrate. Ground Water loo, Waterloo, ON, Canada. Monit. Remediat. 28, 65–72. Volokita, M., Belkin, S., Abeliovich, A., Soares, M.I.M., 1996a. Biological denitrification Robertson, W.D., Ptacek, C.J., Brown, S.J., 2009. Rates of nitrate and perchlorate of drinking water using newspaper. Water Res. 30, 965–971. removal in a 5-year-old wood particle reactor treating agricultural drainage. Volokita, M., Abeliovich, A., Soares, M.I.M., 1996b. Denitrification of groundwater Ground Water Monit. Remediat. 29 (2), 87–94. using cotton as energy source. Water Sci. Technol. 34, 379–385. Saleh, A., Osei, E., Jaynes, D.B., Du, B., Arnold, J.G., 2007. Economic and environmen- Wakatsuki, T., Esumi, H., Omura, S., 1993. High performance and N & P-removable tal impacts of selected BMP’s for nitrate-nitrogen reduction in Walnut Creek on-site domestic waste water treatment system by multi-soil-layer method. Watershed, Iowa, using FEM (farm economic model) and enhanced SWAT (soil Water Sci. Technol. 27, 31–40. and water assessment tool) models. Trans. ASABE 50, 1251–1259. Wallenstein, M.D., Myrold, D.D., Firestone, M., Voytek, M., 2006. Environmental Schipper, L.A., McGill, A., 2008. Nitrogen transformation in a denitrification layer controls on denitrifying communities and denitrification rates: insights from irrigated with dairy factory effluent. Water Res. 42, 2457–2464. molecular methods. Ecol. Appl. 16, 2143–2152. Schipper, L., Vojvodic-Vukovic, M., 1998. Nitrate removal from ground water using Woli, K.P., David, M.B., Cooke, R.A., McIsaac, G.F., Mitchell, C.A., this a denitrification wall amended with sawdust: field trial. J. Environ. Qual. 27, issue. Nitrogen balance in and export from agricultural fields associ- 664–668. ated with controlled drainage and denitrifying bioreactors. Ecol. Eng., Schipper, L.A., Vojvodic-Vukovic, M., 2000. Rates of nitrate removal from ground- doi:10.1016/j.ecoleng.2010.04.024. water and denitrification in a constructed denitrification wall. Ecol. Eng. 14, 269–278.